{
  "id": 8142281,
  "title": "Introducing System One Models & Jev",
  "url": "https://urgent.news/2026/09/18/introducing-system-one-models-jev",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-18T00:05:10.000Z",
  "source": {
    "name": "Lobsters",
    "slug": "lobsters",
    "url": "https://typesafe.ai/blog/introducing-system-one-models-and-jev"
  },
  "original_language": "en",
  "account": "OpenAI's TypeSafe AI has unveiled its first System One Model, Jev, aimed at enabling fast, structured decision-making for software applications. Jev boasts remarkable speed and efficiency, being two orders of magnitude faster than existing language models (LLMs) and achieving comparable levels of intelligence on specialized tasks. Unlike traditional LLMs, Jev doesn't generate strings but instead focuses on structured outputs, eliminating the risk of hallucinations.\n\nJev's unique architecture includes a new model design, a parallel sampler for maximum efficiency, and a training method called Reinforcement Learning for Calibrated Decisions (RLCD). These innovations enable Jev to provide unambiguous, deterministic results, even in complex workflows that require nuanced decision-making.\n\nTypeSafe AI has conducted extensive evaluations to demonstrate Jev's superior performance. The company compares Jev's results to those of GPT-5.6 Terra and other leading models, finding Jev to be significantly faster and more accurate. Notably, Jev's side-by-side demo highlights its ability to output probabilities in parallel, providing more transparency and reliability than autoregressive generation methods used by LLMs.\n\nTo ensure Jev's structured decision-making capabilities, TypeSafe AI has developed a suite of workflows that showcase its strengths. These workflows demonstrate Jev's ability to handle complex, real-world tasks with multiple independent questions and fine-grained probabilities. While some bias may exist due to the evaluation process, the results still suggest Jev's extraordinary potential for automation and business applications.\n\nTypeSafe AI emphasizes the importance of hallucination-free models for automation, arguing that even minor hallucinations can be disastrous in latency-sensitive systems. By focusing on structured decisions and probability-based outputs, Jev aims to provide a more reliable foundation for advanced AI-driven automation.",
  "summary": null,
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}